4 * Copyright (C) 1991, 1992 Linus Torvalds
6 * proc base directory handling functions
8 * 1999, Al Viro. Rewritten. Now it covers the whole per-process part.
9 * Instead of using magical inumbers to determine the kind of object
10 * we allocate and fill in-core inodes upon lookup. They don't even
11 * go into icache. We cache the reference to task_struct upon lookup too.
12 * Eventually it should become a filesystem in its own. We don't use the
13 * rest of procfs anymore.
19 * Bruna Moreira <bruna.moreira@indt.org.br>
20 * Edjard Mota <edjard.mota@indt.org.br>
21 * Ilias Biris <ilias.biris@indt.org.br>
22 * Mauricio Lin <mauricio.lin@indt.org.br>
24 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
26 * A new process specific entry (smaps) included in /proc. It shows the
27 * size of rss for each memory area. The maps entry lacks information
28 * about physical memory size (rss) for each mapped file, i.e.,
29 * rss information for executables and library files.
30 * This additional information is useful for any tools that need to know
31 * about physical memory consumption for a process specific library.
35 * Embedded Linux Lab - 10LE Instituto Nokia de Tecnologia - INdT
36 * Pud inclusion in the page table walking.
40 * 10LE Instituto Nokia de Tecnologia - INdT:
41 * A better way to walks through the page table as suggested by Hugh Dickins.
43 * Simo Piiroinen <simo.piiroinen@nokia.com>:
44 * Smaps information related to shared, private, clean and dirty pages.
46 * Paul Mundt <paul.mundt@nokia.com>:
47 * Overall revision about smaps.
50 #include <asm/uaccess.h>
52 #include <linux/errno.h>
53 #include <linux/time.h>
54 #include <linux/proc_fs.h>
55 #include <linux/stat.h>
56 #include <linux/task_io_accounting_ops.h>
57 #include <linux/init.h>
58 #include <linux/capability.h>
59 #include <linux/file.h>
60 #include <linux/fdtable.h>
61 #include <linux/string.h>
62 #include <linux/seq_file.h>
63 #include <linux/namei.h>
64 #include <linux/mnt_namespace.h>
66 #include <linux/swap.h>
67 #include <linux/rcupdate.h>
68 #include <linux/kallsyms.h>
69 #include <linux/stacktrace.h>
70 #include <linux/resource.h>
71 #include <linux/module.h>
72 #include <linux/mount.h>
73 #include <linux/security.h>
74 #include <linux/ptrace.h>
75 #include <linux/tracehook.h>
76 #include <linux/cgroup.h>
77 #include <linux/cpuset.h>
78 #include <linux/audit.h>
79 #include <linux/poll.h>
80 #include <linux/nsproxy.h>
81 #include <linux/oom.h>
82 #include <linux/elf.h>
83 #include <linux/pid_namespace.h>
84 #include <linux/user_namespace.h>
85 #include <linux/fs_struct.h>
86 #include <linux/slab.h>
87 #include <linux/flex_array.h>
88 #ifdef CONFIG_HARDWALL
89 #include <asm/hardwall.h>
91 #include <trace/events/oom.h>
95 * Implementing inode permission operations in /proc is almost
96 * certainly an error. Permission checks need to happen during
97 * each system call not at open time. The reason is that most of
98 * what we wish to check for permissions in /proc varies at runtime.
100 * The classic example of a problem is opening file descriptors
101 * in /proc for a task before it execs a suid executable.
108 const struct inode_operations *iop;
109 const struct file_operations *fop;
113 #define NOD(NAME, MODE, IOP, FOP, OP) { \
115 .len = sizeof(NAME) - 1, \
122 #define DIR(NAME, MODE, iops, fops) \
123 NOD(NAME, (S_IFDIR|(MODE)), &iops, &fops, {} )
124 #define LNK(NAME, get_link) \
125 NOD(NAME, (S_IFLNK|S_IRWXUGO), \
126 &proc_pid_link_inode_operations, NULL, \
127 { .proc_get_link = get_link } )
128 #define REG(NAME, MODE, fops) \
129 NOD(NAME, (S_IFREG|(MODE)), NULL, &fops, {})
130 #define INF(NAME, MODE, read) \
131 NOD(NAME, (S_IFREG|(MODE)), \
132 NULL, &proc_info_file_operations, \
133 { .proc_read = read } )
134 #define ONE(NAME, MODE, show) \
135 NOD(NAME, (S_IFREG|(MODE)), \
136 NULL, &proc_single_file_operations, \
137 { .proc_show = show } )
139 static int proc_fd_permission(struct inode *inode, int mask);
142 * Count the number of hardlinks for the pid_entry table, excluding the .
145 static unsigned int pid_entry_count_dirs(const struct pid_entry *entries,
152 for (i = 0; i < n; ++i) {
153 if (S_ISDIR(entries[i].mode))
160 static int get_task_root(struct task_struct *task, struct path *root)
162 int result = -ENOENT;
166 get_fs_root(task->fs, root);
173 static int proc_cwd_link(struct dentry *dentry, struct path *path)
175 struct task_struct *task = get_proc_task(dentry->d_inode);
176 int result = -ENOENT;
181 get_fs_pwd(task->fs, path);
185 put_task_struct(task);
190 static int proc_root_link(struct dentry *dentry, struct path *path)
192 struct task_struct *task = get_proc_task(dentry->d_inode);
193 int result = -ENOENT;
196 result = get_task_root(task, path);
197 put_task_struct(task);
202 struct mm_struct *mm_for_maps(struct task_struct *task)
204 return mm_access(task, PTRACE_MODE_READ);
207 static int proc_pid_cmdline(struct task_struct *task, char * buffer)
211 struct mm_struct *mm = get_task_mm(task);
215 goto out_mm; /* Shh! No looking before we're done */
217 len = mm->arg_end - mm->arg_start;
222 res = access_process_vm(task, mm->arg_start, buffer, len, 0);
224 // If the nul at the end of args has been overwritten, then
225 // assume application is using setproctitle(3).
226 if (res > 0 && buffer[res-1] != '\0' && len < PAGE_SIZE) {
227 len = strnlen(buffer, res);
231 len = mm->env_end - mm->env_start;
232 if (len > PAGE_SIZE - res)
233 len = PAGE_SIZE - res;
234 res += access_process_vm(task, mm->env_start, buffer+res, len, 0);
235 res = strnlen(buffer, res);
244 static int proc_pid_auxv(struct task_struct *task, char *buffer)
246 struct mm_struct *mm = mm_for_maps(task);
247 int res = PTR_ERR(mm);
248 if (mm && !IS_ERR(mm)) {
249 unsigned int nwords = 0;
252 } while (mm->saved_auxv[nwords - 2] != 0); /* AT_NULL */
253 res = nwords * sizeof(mm->saved_auxv[0]);
256 memcpy(buffer, mm->saved_auxv, res);
263 #ifdef CONFIG_KALLSYMS
265 * Provides a wchan file via kallsyms in a proper one-value-per-file format.
266 * Returns the resolved symbol. If that fails, simply return the address.
268 static int proc_pid_wchan(struct task_struct *task, char *buffer)
271 char symname[KSYM_NAME_LEN];
273 wchan = get_wchan(task);
275 if (lookup_symbol_name(wchan, symname) < 0)
276 if (!ptrace_may_access(task, PTRACE_MODE_READ))
279 return sprintf(buffer, "%lu", wchan);
281 return sprintf(buffer, "%s", symname);
283 #endif /* CONFIG_KALLSYMS */
285 static int lock_trace(struct task_struct *task)
287 int err = mutex_lock_killable(&task->signal->cred_guard_mutex);
290 if (!ptrace_may_access(task, PTRACE_MODE_ATTACH)) {
291 mutex_unlock(&task->signal->cred_guard_mutex);
297 static void unlock_trace(struct task_struct *task)
299 mutex_unlock(&task->signal->cred_guard_mutex);
302 #ifdef CONFIG_STACKTRACE
304 #define MAX_STACK_TRACE_DEPTH 64
306 static int proc_pid_stack(struct seq_file *m, struct pid_namespace *ns,
307 struct pid *pid, struct task_struct *task)
309 struct stack_trace trace;
310 unsigned long *entries;
314 entries = kmalloc(MAX_STACK_TRACE_DEPTH * sizeof(*entries), GFP_KERNEL);
318 trace.nr_entries = 0;
319 trace.max_entries = MAX_STACK_TRACE_DEPTH;
320 trace.entries = entries;
323 err = lock_trace(task);
325 save_stack_trace_tsk(task, &trace);
327 for (i = 0; i < trace.nr_entries; i++) {
328 seq_printf(m, "[<%pK>] %pS\n",
329 (void *)entries[i], (void *)entries[i]);
339 #ifdef CONFIG_SCHEDSTATS
341 * Provides /proc/PID/schedstat
343 static int proc_pid_schedstat(struct task_struct *task, char *buffer)
345 return sprintf(buffer, "%llu %llu %lu\n",
346 (unsigned long long)task->se.sum_exec_runtime,
347 (unsigned long long)task->sched_info.run_delay,
348 task->sched_info.pcount);
352 #ifdef CONFIG_LATENCYTOP
353 static int lstats_show_proc(struct seq_file *m, void *v)
356 struct inode *inode = m->private;
357 struct task_struct *task = get_proc_task(inode);
361 seq_puts(m, "Latency Top version : v0.1\n");
362 for (i = 0; i < 32; i++) {
363 struct latency_record *lr = &task->latency_record[i];
364 if (lr->backtrace[0]) {
366 seq_printf(m, "%i %li %li",
367 lr->count, lr->time, lr->max);
368 for (q = 0; q < LT_BACKTRACEDEPTH; q++) {
369 unsigned long bt = lr->backtrace[q];
374 seq_printf(m, " %ps", (void *)bt);
380 put_task_struct(task);
384 static int lstats_open(struct inode *inode, struct file *file)
386 return single_open(file, lstats_show_proc, inode);
389 static ssize_t lstats_write(struct file *file, const char __user *buf,
390 size_t count, loff_t *offs)
392 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
396 clear_all_latency_tracing(task);
397 put_task_struct(task);
402 static const struct file_operations proc_lstats_operations = {
405 .write = lstats_write,
407 .release = single_release,
412 static int proc_oom_score(struct task_struct *task, char *buffer)
414 unsigned long totalpages = totalram_pages + total_swap_pages;
415 unsigned long points = 0;
417 read_lock(&tasklist_lock);
419 points = oom_badness(task, NULL, NULL, totalpages) *
421 read_unlock(&tasklist_lock);
422 return sprintf(buffer, "%lu\n", points);
430 static const struct limit_names lnames[RLIM_NLIMITS] = {
431 [RLIMIT_CPU] = {"Max cpu time", "seconds"},
432 [RLIMIT_FSIZE] = {"Max file size", "bytes"},
433 [RLIMIT_DATA] = {"Max data size", "bytes"},
434 [RLIMIT_STACK] = {"Max stack size", "bytes"},
435 [RLIMIT_CORE] = {"Max core file size", "bytes"},
436 [RLIMIT_RSS] = {"Max resident set", "bytes"},
437 [RLIMIT_NPROC] = {"Max processes", "processes"},
438 [RLIMIT_NOFILE] = {"Max open files", "files"},
439 [RLIMIT_MEMLOCK] = {"Max locked memory", "bytes"},
440 [RLIMIT_AS] = {"Max address space", "bytes"},
441 [RLIMIT_LOCKS] = {"Max file locks", "locks"},
442 [RLIMIT_SIGPENDING] = {"Max pending signals", "signals"},
443 [RLIMIT_MSGQUEUE] = {"Max msgqueue size", "bytes"},
444 [RLIMIT_NICE] = {"Max nice priority", NULL},
445 [RLIMIT_RTPRIO] = {"Max realtime priority", NULL},
446 [RLIMIT_RTTIME] = {"Max realtime timeout", "us"},
449 /* Display limits for a process */
450 static int proc_pid_limits(struct task_struct *task, char *buffer)
455 char *bufptr = buffer;
457 struct rlimit rlim[RLIM_NLIMITS];
459 if (!lock_task_sighand(task, &flags))
461 memcpy(rlim, task->signal->rlim, sizeof(struct rlimit) * RLIM_NLIMITS);
462 unlock_task_sighand(task, &flags);
465 * print the file header
467 count += sprintf(&bufptr[count], "%-25s %-20s %-20s %-10s\n",
468 "Limit", "Soft Limit", "Hard Limit", "Units");
470 for (i = 0; i < RLIM_NLIMITS; i++) {
471 if (rlim[i].rlim_cur == RLIM_INFINITY)
472 count += sprintf(&bufptr[count], "%-25s %-20s ",
473 lnames[i].name, "unlimited");
475 count += sprintf(&bufptr[count], "%-25s %-20lu ",
476 lnames[i].name, rlim[i].rlim_cur);
478 if (rlim[i].rlim_max == RLIM_INFINITY)
479 count += sprintf(&bufptr[count], "%-20s ", "unlimited");
481 count += sprintf(&bufptr[count], "%-20lu ",
485 count += sprintf(&bufptr[count], "%-10s\n",
488 count += sprintf(&bufptr[count], "\n");
494 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
495 static int proc_pid_syscall(struct task_struct *task, char *buffer)
498 unsigned long args[6], sp, pc;
499 int res = lock_trace(task);
503 if (task_current_syscall(task, &nr, args, 6, &sp, &pc))
504 res = sprintf(buffer, "running\n");
506 res = sprintf(buffer, "%ld 0x%lx 0x%lx\n", nr, sp, pc);
508 res = sprintf(buffer,
509 "%ld 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx 0x%lx\n",
511 args[0], args[1], args[2], args[3], args[4], args[5],
516 #endif /* CONFIG_HAVE_ARCH_TRACEHOOK */
518 /************************************************************************/
519 /* Here the fs part begins */
520 /************************************************************************/
522 /* permission checks */
523 static int proc_fd_access_allowed(struct inode *inode)
525 struct task_struct *task;
527 /* Allow access to a task's file descriptors if it is us or we
528 * may use ptrace attach to the process and find out that
531 task = get_proc_task(inode);
533 allowed = ptrace_may_access(task, PTRACE_MODE_READ);
534 put_task_struct(task);
539 int proc_setattr(struct dentry *dentry, struct iattr *attr)
542 struct inode *inode = dentry->d_inode;
544 if (attr->ia_valid & ATTR_MODE)
547 error = inode_change_ok(inode, attr);
551 if ((attr->ia_valid & ATTR_SIZE) &&
552 attr->ia_size != i_size_read(inode)) {
553 error = vmtruncate(inode, attr->ia_size);
558 setattr_copy(inode, attr);
559 mark_inode_dirty(inode);
564 * May current process learn task's sched/cmdline info (for hide_pid_min=1)
565 * or euid/egid (for hide_pid_min=2)?
567 static bool has_pid_permissions(struct pid_namespace *pid,
568 struct task_struct *task,
571 if (pid->hide_pid < hide_pid_min)
573 if (in_group_p(pid->pid_gid))
575 return ptrace_may_access(task, PTRACE_MODE_READ);
579 static int proc_pid_permission(struct inode *inode, int mask)
581 struct pid_namespace *pid = inode->i_sb->s_fs_info;
582 struct task_struct *task;
585 task = get_proc_task(inode);
588 has_perms = has_pid_permissions(pid, task, 1);
589 put_task_struct(task);
592 if (pid->hide_pid == 2) {
594 * Let's make getdents(), stat(), and open()
595 * consistent with each other. If a process
596 * may not stat() a file, it shouldn't be seen
604 return generic_permission(inode, mask);
609 static const struct inode_operations proc_def_inode_operations = {
610 .setattr = proc_setattr,
613 #define PROC_BLOCK_SIZE (3*1024) /* 4K page size but our output routines use some slack for overruns */
615 static ssize_t proc_info_read(struct file * file, char __user * buf,
616 size_t count, loff_t *ppos)
618 struct inode * inode = file->f_path.dentry->d_inode;
621 struct task_struct *task = get_proc_task(inode);
627 if (count > PROC_BLOCK_SIZE)
628 count = PROC_BLOCK_SIZE;
631 if (!(page = __get_free_page(GFP_TEMPORARY)))
634 length = PROC_I(inode)->op.proc_read(task, (char*)page);
637 length = simple_read_from_buffer(buf, count, ppos, (char *)page, length);
640 put_task_struct(task);
645 static const struct file_operations proc_info_file_operations = {
646 .read = proc_info_read,
647 .llseek = generic_file_llseek,
650 static int proc_single_show(struct seq_file *m, void *v)
652 struct inode *inode = m->private;
653 struct pid_namespace *ns;
655 struct task_struct *task;
658 ns = inode->i_sb->s_fs_info;
659 pid = proc_pid(inode);
660 task = get_pid_task(pid, PIDTYPE_PID);
664 ret = PROC_I(inode)->op.proc_show(m, ns, pid, task);
666 put_task_struct(task);
670 static int proc_single_open(struct inode *inode, struct file *filp)
672 return single_open(filp, proc_single_show, inode);
675 static const struct file_operations proc_single_file_operations = {
676 .open = proc_single_open,
679 .release = single_release,
682 static int __mem_open(struct inode *inode, struct file *file, unsigned int mode)
684 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
685 struct mm_struct *mm;
690 mm = mm_access(task, mode);
691 put_task_struct(task);
697 /* ensure this mm_struct can't be freed */
698 atomic_inc(&mm->mm_count);
699 /* but do not pin its memory */
703 /* OK to pass negative loff_t, we can catch out-of-range */
704 file->f_mode |= FMODE_UNSIGNED_OFFSET;
705 file->private_data = mm;
710 static int mem_open(struct inode *inode, struct file *file)
712 return __mem_open(inode, file, PTRACE_MODE_ATTACH);
715 static ssize_t mem_rw(struct file *file, char __user *buf,
716 size_t count, loff_t *ppos, int write)
718 struct mm_struct *mm = file->private_data;
719 unsigned long addr = *ppos;
726 page = (char *)__get_free_page(GFP_TEMPORARY);
731 if (!atomic_inc_not_zero(&mm->mm_users))
735 int this_len = min_t(int, count, PAGE_SIZE);
737 if (write && copy_from_user(page, buf, this_len)) {
742 this_len = access_remote_vm(mm, addr, page, this_len, write);
749 if (!write && copy_to_user(buf, page, this_len)) {
763 free_page((unsigned long) page);
767 static ssize_t mem_read(struct file *file, char __user *buf,
768 size_t count, loff_t *ppos)
770 return mem_rw(file, buf, count, ppos, 0);
773 static ssize_t mem_write(struct file *file, const char __user *buf,
774 size_t count, loff_t *ppos)
776 return mem_rw(file, (char __user*)buf, count, ppos, 1);
779 loff_t mem_lseek(struct file *file, loff_t offset, int orig)
783 file->f_pos = offset;
786 file->f_pos += offset;
791 force_successful_syscall_return();
795 static int mem_release(struct inode *inode, struct file *file)
797 struct mm_struct *mm = file->private_data;
803 static const struct file_operations proc_mem_operations = {
808 .release = mem_release,
811 static int environ_open(struct inode *inode, struct file *file)
813 return __mem_open(inode, file, PTRACE_MODE_READ);
816 static ssize_t environ_read(struct file *file, char __user *buf,
817 size_t count, loff_t *ppos)
820 unsigned long src = *ppos;
822 struct mm_struct *mm = file->private_data;
827 page = (char *)__get_free_page(GFP_TEMPORARY);
832 if (!atomic_inc_not_zero(&mm->mm_users))
835 int this_len, retval, max_len;
837 this_len = mm->env_end - (mm->env_start + src);
842 max_len = (count > PAGE_SIZE) ? PAGE_SIZE : count;
843 this_len = (this_len > max_len) ? max_len : this_len;
845 retval = access_remote_vm(mm, (mm->env_start + src),
853 if (copy_to_user(buf, page, retval)) {
867 free_page((unsigned long) page);
871 static const struct file_operations proc_environ_operations = {
872 .open = environ_open,
873 .read = environ_read,
874 .llseek = generic_file_llseek,
875 .release = mem_release,
878 static ssize_t oom_adjust_read(struct file *file, char __user *buf,
879 size_t count, loff_t *ppos)
881 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
882 char buffer[PROC_NUMBUF];
884 int oom_adjust = OOM_DISABLE;
890 if (lock_task_sighand(task, &flags)) {
891 oom_adjust = task->signal->oom_adj;
892 unlock_task_sighand(task, &flags);
895 put_task_struct(task);
897 len = snprintf(buffer, sizeof(buffer), "%i\n", oom_adjust);
899 return simple_read_from_buffer(buf, count, ppos, buffer, len);
902 static ssize_t oom_adjust_write(struct file *file, const char __user *buf,
903 size_t count, loff_t *ppos)
905 struct task_struct *task;
906 char buffer[PROC_NUMBUF];
911 memset(buffer, 0, sizeof(buffer));
912 if (count > sizeof(buffer) - 1)
913 count = sizeof(buffer) - 1;
914 if (copy_from_user(buffer, buf, count)) {
919 err = kstrtoint(strstrip(buffer), 0, &oom_adjust);
922 if ((oom_adjust < OOM_ADJUST_MIN || oom_adjust > OOM_ADJUST_MAX) &&
923 oom_adjust != OOM_DISABLE) {
928 task = get_proc_task(file->f_path.dentry->d_inode);
940 if (!lock_task_sighand(task, &flags)) {
945 if (oom_adjust < task->signal->oom_adj && !capable(CAP_SYS_RESOURCE)) {
951 * Warn that /proc/pid/oom_adj is deprecated, see
952 * Documentation/feature-removal-schedule.txt.
954 printk_once(KERN_WARNING "%s (%d): /proc/%d/oom_adj is deprecated, please use /proc/%d/oom_score_adj instead.\n",
955 current->comm, task_pid_nr(current), task_pid_nr(task),
957 task->signal->oom_adj = oom_adjust;
959 * Scale /proc/pid/oom_score_adj appropriately ensuring that a maximum
960 * value is always attainable.
962 if (task->signal->oom_adj == OOM_ADJUST_MAX)
963 task->signal->oom_score_adj = OOM_SCORE_ADJ_MAX;
965 task->signal->oom_score_adj = (oom_adjust * OOM_SCORE_ADJ_MAX) /
967 trace_oom_score_adj_update(task);
969 unlock_task_sighand(task, &flags);
972 put_task_struct(task);
974 return err < 0 ? err : count;
977 static const struct file_operations proc_oom_adjust_operations = {
978 .read = oom_adjust_read,
979 .write = oom_adjust_write,
980 .llseek = generic_file_llseek,
983 static ssize_t oom_score_adj_read(struct file *file, char __user *buf,
984 size_t count, loff_t *ppos)
986 struct task_struct *task = get_proc_task(file->f_path.dentry->d_inode);
987 char buffer[PROC_NUMBUF];
988 int oom_score_adj = OOM_SCORE_ADJ_MIN;
994 if (lock_task_sighand(task, &flags)) {
995 oom_score_adj = task->signal->oom_score_adj;
996 unlock_task_sighand(task, &flags);
998 put_task_struct(task);
999 len = snprintf(buffer, sizeof(buffer), "%d\n", oom_score_adj);
1000 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1003 static ssize_t oom_score_adj_write(struct file *file, const char __user *buf,
1004 size_t count, loff_t *ppos)
1006 struct task_struct *task;
1007 char buffer[PROC_NUMBUF];
1008 unsigned long flags;
1012 memset(buffer, 0, sizeof(buffer));
1013 if (count > sizeof(buffer) - 1)
1014 count = sizeof(buffer) - 1;
1015 if (copy_from_user(buffer, buf, count)) {
1020 err = kstrtoint(strstrip(buffer), 0, &oom_score_adj);
1023 if (oom_score_adj < OOM_SCORE_ADJ_MIN ||
1024 oom_score_adj > OOM_SCORE_ADJ_MAX) {
1029 task = get_proc_task(file->f_path.dentry->d_inode);
1041 if (!lock_task_sighand(task, &flags)) {
1046 if (oom_score_adj < task->signal->oom_score_adj_min &&
1047 !capable(CAP_SYS_RESOURCE)) {
1052 task->signal->oom_score_adj = oom_score_adj;
1053 if (has_capability_noaudit(current, CAP_SYS_RESOURCE))
1054 task->signal->oom_score_adj_min = oom_score_adj;
1055 trace_oom_score_adj_update(task);
1057 * Scale /proc/pid/oom_adj appropriately ensuring that OOM_DISABLE is
1058 * always attainable.
1060 if (task->signal->oom_score_adj == OOM_SCORE_ADJ_MIN)
1061 task->signal->oom_adj = OOM_DISABLE;
1063 task->signal->oom_adj = (oom_score_adj * OOM_ADJUST_MAX) /
1066 unlock_task_sighand(task, &flags);
1069 put_task_struct(task);
1071 return err < 0 ? err : count;
1074 static const struct file_operations proc_oom_score_adj_operations = {
1075 .read = oom_score_adj_read,
1076 .write = oom_score_adj_write,
1077 .llseek = default_llseek,
1080 #ifdef CONFIG_AUDITSYSCALL
1081 #define TMPBUFLEN 21
1082 static ssize_t proc_loginuid_read(struct file * file, char __user * buf,
1083 size_t count, loff_t *ppos)
1085 struct inode * inode = file->f_path.dentry->d_inode;
1086 struct task_struct *task = get_proc_task(inode);
1088 char tmpbuf[TMPBUFLEN];
1092 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1093 audit_get_loginuid(task));
1094 put_task_struct(task);
1095 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1098 static ssize_t proc_loginuid_write(struct file * file, const char __user * buf,
1099 size_t count, loff_t *ppos)
1101 struct inode * inode = file->f_path.dentry->d_inode;
1107 if (current != pid_task(proc_pid(inode), PIDTYPE_PID)) {
1113 if (count >= PAGE_SIZE)
1114 count = PAGE_SIZE - 1;
1117 /* No partial writes. */
1120 page = (char*)__get_free_page(GFP_TEMPORARY);
1124 if (copy_from_user(page, buf, count))
1128 loginuid = simple_strtoul(page, &tmp, 10);
1134 length = audit_set_loginuid(loginuid);
1135 if (likely(length == 0))
1139 free_page((unsigned long) page);
1143 static const struct file_operations proc_loginuid_operations = {
1144 .read = proc_loginuid_read,
1145 .write = proc_loginuid_write,
1146 .llseek = generic_file_llseek,
1149 static ssize_t proc_sessionid_read(struct file * file, char __user * buf,
1150 size_t count, loff_t *ppos)
1152 struct inode * inode = file->f_path.dentry->d_inode;
1153 struct task_struct *task = get_proc_task(inode);
1155 char tmpbuf[TMPBUFLEN];
1159 length = scnprintf(tmpbuf, TMPBUFLEN, "%u",
1160 audit_get_sessionid(task));
1161 put_task_struct(task);
1162 return simple_read_from_buffer(buf, count, ppos, tmpbuf, length);
1165 static const struct file_operations proc_sessionid_operations = {
1166 .read = proc_sessionid_read,
1167 .llseek = generic_file_llseek,
1171 #ifdef CONFIG_FAULT_INJECTION
1172 static ssize_t proc_fault_inject_read(struct file * file, char __user * buf,
1173 size_t count, loff_t *ppos)
1175 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
1176 char buffer[PROC_NUMBUF];
1182 make_it_fail = task->make_it_fail;
1183 put_task_struct(task);
1185 len = snprintf(buffer, sizeof(buffer), "%i\n", make_it_fail);
1187 return simple_read_from_buffer(buf, count, ppos, buffer, len);
1190 static ssize_t proc_fault_inject_write(struct file * file,
1191 const char __user * buf, size_t count, loff_t *ppos)
1193 struct task_struct *task;
1194 char buffer[PROC_NUMBUF], *end;
1197 if (!capable(CAP_SYS_RESOURCE))
1199 memset(buffer, 0, sizeof(buffer));
1200 if (count > sizeof(buffer) - 1)
1201 count = sizeof(buffer) - 1;
1202 if (copy_from_user(buffer, buf, count))
1204 make_it_fail = simple_strtol(strstrip(buffer), &end, 0);
1207 task = get_proc_task(file->f_dentry->d_inode);
1210 task->make_it_fail = make_it_fail;
1211 put_task_struct(task);
1216 static const struct file_operations proc_fault_inject_operations = {
1217 .read = proc_fault_inject_read,
1218 .write = proc_fault_inject_write,
1219 .llseek = generic_file_llseek,
1224 #ifdef CONFIG_SCHED_DEBUG
1226 * Print out various scheduling related per-task fields:
1228 static int sched_show(struct seq_file *m, void *v)
1230 struct inode *inode = m->private;
1231 struct task_struct *p;
1233 p = get_proc_task(inode);
1236 proc_sched_show_task(p, m);
1244 sched_write(struct file *file, const char __user *buf,
1245 size_t count, loff_t *offset)
1247 struct inode *inode = file->f_path.dentry->d_inode;
1248 struct task_struct *p;
1250 p = get_proc_task(inode);
1253 proc_sched_set_task(p);
1260 static int sched_open(struct inode *inode, struct file *filp)
1262 return single_open(filp, sched_show, inode);
1265 static const struct file_operations proc_pid_sched_operations = {
1268 .write = sched_write,
1269 .llseek = seq_lseek,
1270 .release = single_release,
1275 #ifdef CONFIG_SCHED_AUTOGROUP
1277 * Print out autogroup related information:
1279 static int sched_autogroup_show(struct seq_file *m, void *v)
1281 struct inode *inode = m->private;
1282 struct task_struct *p;
1284 p = get_proc_task(inode);
1287 proc_sched_autogroup_show_task(p, m);
1295 sched_autogroup_write(struct file *file, const char __user *buf,
1296 size_t count, loff_t *offset)
1298 struct inode *inode = file->f_path.dentry->d_inode;
1299 struct task_struct *p;
1300 char buffer[PROC_NUMBUF];
1304 memset(buffer, 0, sizeof(buffer));
1305 if (count > sizeof(buffer) - 1)
1306 count = sizeof(buffer) - 1;
1307 if (copy_from_user(buffer, buf, count))
1310 err = kstrtoint(strstrip(buffer), 0, &nice);
1314 p = get_proc_task(inode);
1318 err = proc_sched_autogroup_set_nice(p, nice);
1327 static int sched_autogroup_open(struct inode *inode, struct file *filp)
1331 ret = single_open(filp, sched_autogroup_show, NULL);
1333 struct seq_file *m = filp->private_data;
1340 static const struct file_operations proc_pid_sched_autogroup_operations = {
1341 .open = sched_autogroup_open,
1343 .write = sched_autogroup_write,
1344 .llseek = seq_lseek,
1345 .release = single_release,
1348 #endif /* CONFIG_SCHED_AUTOGROUP */
1350 static ssize_t comm_write(struct file *file, const char __user *buf,
1351 size_t count, loff_t *offset)
1353 struct inode *inode = file->f_path.dentry->d_inode;
1354 struct task_struct *p;
1355 char buffer[TASK_COMM_LEN];
1357 memset(buffer, 0, sizeof(buffer));
1358 if (count > sizeof(buffer) - 1)
1359 count = sizeof(buffer) - 1;
1360 if (copy_from_user(buffer, buf, count))
1363 p = get_proc_task(inode);
1367 if (same_thread_group(current, p))
1368 set_task_comm(p, buffer);
1377 static int comm_show(struct seq_file *m, void *v)
1379 struct inode *inode = m->private;
1380 struct task_struct *p;
1382 p = get_proc_task(inode);
1387 seq_printf(m, "%s\n", p->comm);
1395 static int comm_open(struct inode *inode, struct file *filp)
1397 return single_open(filp, comm_show, inode);
1400 static const struct file_operations proc_pid_set_comm_operations = {
1403 .write = comm_write,
1404 .llseek = seq_lseek,
1405 .release = single_release,
1408 static int proc_exe_link(struct dentry *dentry, struct path *exe_path)
1410 struct task_struct *task;
1411 struct mm_struct *mm;
1412 struct file *exe_file;
1414 task = get_proc_task(dentry->d_inode);
1417 mm = get_task_mm(task);
1418 put_task_struct(task);
1421 exe_file = get_mm_exe_file(mm);
1424 *exe_path = exe_file->f_path;
1425 path_get(&exe_file->f_path);
1432 static void *proc_pid_follow_link(struct dentry *dentry, struct nameidata *nd)
1434 struct inode *inode = dentry->d_inode;
1435 int error = -EACCES;
1437 /* We don't need a base pointer in the /proc filesystem */
1438 path_put(&nd->path);
1440 /* Are we allowed to snoop on the tasks file descriptors? */
1441 if (!proc_fd_access_allowed(inode))
1444 error = PROC_I(inode)->op.proc_get_link(dentry, &nd->path);
1446 return ERR_PTR(error);
1449 static int do_proc_readlink(struct path *path, char __user *buffer, int buflen)
1451 char *tmp = (char*)__get_free_page(GFP_TEMPORARY);
1458 pathname = d_path(path, tmp, PAGE_SIZE);
1459 len = PTR_ERR(pathname);
1460 if (IS_ERR(pathname))
1462 len = tmp + PAGE_SIZE - 1 - pathname;
1466 if (copy_to_user(buffer, pathname, len))
1469 free_page((unsigned long)tmp);
1473 static int proc_pid_readlink(struct dentry * dentry, char __user * buffer, int buflen)
1475 int error = -EACCES;
1476 struct inode *inode = dentry->d_inode;
1479 /* Are we allowed to snoop on the tasks file descriptors? */
1480 if (!proc_fd_access_allowed(inode))
1483 error = PROC_I(inode)->op.proc_get_link(dentry, &path);
1487 error = do_proc_readlink(&path, buffer, buflen);
1493 static const struct inode_operations proc_pid_link_inode_operations = {
1494 .readlink = proc_pid_readlink,
1495 .follow_link = proc_pid_follow_link,
1496 .setattr = proc_setattr,
1500 /* building an inode */
1502 static int task_dumpable(struct task_struct *task)
1505 struct mm_struct *mm;
1510 dumpable = get_dumpable(mm);
1517 struct inode *proc_pid_make_inode(struct super_block * sb, struct task_struct *task)
1519 struct inode * inode;
1520 struct proc_inode *ei;
1521 const struct cred *cred;
1523 /* We need a new inode */
1525 inode = new_inode(sb);
1531 inode->i_ino = get_next_ino();
1532 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
1533 inode->i_op = &proc_def_inode_operations;
1536 * grab the reference to task.
1538 ei->pid = get_task_pid(task, PIDTYPE_PID);
1542 if (task_dumpable(task)) {
1544 cred = __task_cred(task);
1545 inode->i_uid = cred->euid;
1546 inode->i_gid = cred->egid;
1549 security_task_to_inode(task, inode);
1559 int pid_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
1561 struct inode *inode = dentry->d_inode;
1562 struct task_struct *task;
1563 const struct cred *cred;
1564 struct pid_namespace *pid = dentry->d_sb->s_fs_info;
1566 generic_fillattr(inode, stat);
1569 stat->uid = GLOBAL_ROOT_UID;
1570 stat->gid = GLOBAL_ROOT_GID;
1571 task = pid_task(proc_pid(inode), PIDTYPE_PID);
1573 if (!has_pid_permissions(pid, task, 2)) {
1576 * This doesn't prevent learning whether PID exists,
1577 * it only makes getattr() consistent with readdir().
1581 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1582 task_dumpable(task)) {
1583 cred = __task_cred(task);
1584 stat->uid = cred->euid;
1585 stat->gid = cred->egid;
1595 * Exceptional case: normally we are not allowed to unhash a busy
1596 * directory. In this case, however, we can do it - no aliasing problems
1597 * due to the way we treat inodes.
1599 * Rewrite the inode's ownerships here because the owning task may have
1600 * performed a setuid(), etc.
1602 * Before the /proc/pid/status file was created the only way to read
1603 * the effective uid of a /process was to stat /proc/pid. Reading
1604 * /proc/pid/status is slow enough that procps and other packages
1605 * kept stating /proc/pid. To keep the rules in /proc simple I have
1606 * made this apply to all per process world readable and executable
1609 int pid_revalidate(struct dentry *dentry, struct nameidata *nd)
1611 struct inode *inode;
1612 struct task_struct *task;
1613 const struct cred *cred;
1615 if (nd && nd->flags & LOOKUP_RCU)
1618 inode = dentry->d_inode;
1619 task = get_proc_task(inode);
1622 if ((inode->i_mode == (S_IFDIR|S_IRUGO|S_IXUGO)) ||
1623 task_dumpable(task)) {
1625 cred = __task_cred(task);
1626 inode->i_uid = cred->euid;
1627 inode->i_gid = cred->egid;
1630 inode->i_uid = GLOBAL_ROOT_UID;
1631 inode->i_gid = GLOBAL_ROOT_GID;
1633 inode->i_mode &= ~(S_ISUID | S_ISGID);
1634 security_task_to_inode(task, inode);
1635 put_task_struct(task);
1642 static int pid_delete_dentry(const struct dentry * dentry)
1644 /* Is the task we represent dead?
1645 * If so, then don't put the dentry on the lru list,
1646 * kill it immediately.
1648 return !proc_pid(dentry->d_inode)->tasks[PIDTYPE_PID].first;
1651 const struct dentry_operations pid_dentry_operations =
1653 .d_revalidate = pid_revalidate,
1654 .d_delete = pid_delete_dentry,
1660 * Fill a directory entry.
1662 * If possible create the dcache entry and derive our inode number and
1663 * file type from dcache entry.
1665 * Since all of the proc inode numbers are dynamically generated, the inode
1666 * numbers do not exist until the inode is cache. This means creating the
1667 * the dcache entry in readdir is necessary to keep the inode numbers
1668 * reported by readdir in sync with the inode numbers reported
1671 int proc_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
1672 const char *name, int len,
1673 instantiate_t instantiate, struct task_struct *task, const void *ptr)
1675 struct dentry *child, *dir = filp->f_path.dentry;
1676 struct inode *inode;
1679 unsigned type = DT_UNKNOWN;
1683 qname.hash = full_name_hash(name, len);
1685 child = d_lookup(dir, &qname);
1688 new = d_alloc(dir, &qname);
1690 child = instantiate(dir->d_inode, new, task, ptr);
1697 if (!child || IS_ERR(child) || !child->d_inode)
1698 goto end_instantiate;
1699 inode = child->d_inode;
1702 type = inode->i_mode >> 12;
1707 ino = find_inode_number(dir, &qname);
1710 return filldir(dirent, name, len, filp->f_pos, ino, type);
1713 static unsigned name_to_int(struct dentry *dentry)
1715 const char *name = dentry->d_name.name;
1716 int len = dentry->d_name.len;
1719 if (len > 1 && *name == '0')
1722 unsigned c = *name++ - '0';
1725 if (n >= (~0U-9)/10)
1735 #define PROC_FDINFO_MAX 64
1737 static int proc_fd_info(struct inode *inode, struct path *path, char *info)
1739 struct task_struct *task = get_proc_task(inode);
1740 struct files_struct *files = NULL;
1742 int fd = proc_fd(inode);
1745 files = get_files_struct(task);
1746 put_task_struct(task);
1750 * We are not taking a ref to the file structure, so we must
1753 spin_lock(&files->file_lock);
1754 file = fcheck_files(files, fd);
1756 unsigned int f_flags;
1757 struct fdtable *fdt;
1759 fdt = files_fdtable(files);
1760 f_flags = file->f_flags & ~O_CLOEXEC;
1761 if (close_on_exec(fd, fdt))
1762 f_flags |= O_CLOEXEC;
1765 *path = file->f_path;
1766 path_get(&file->f_path);
1769 snprintf(info, PROC_FDINFO_MAX,
1772 (long long) file->f_pos,
1774 spin_unlock(&files->file_lock);
1775 put_files_struct(files);
1778 spin_unlock(&files->file_lock);
1779 put_files_struct(files);
1784 static int proc_fd_link(struct dentry *dentry, struct path *path)
1786 return proc_fd_info(dentry->d_inode, path, NULL);
1789 static int tid_fd_revalidate(struct dentry *dentry, struct nameidata *nd)
1791 struct inode *inode;
1792 struct task_struct *task;
1794 struct files_struct *files;
1795 const struct cred *cred;
1797 if (nd && nd->flags & LOOKUP_RCU)
1800 inode = dentry->d_inode;
1801 task = get_proc_task(inode);
1802 fd = proc_fd(inode);
1805 files = get_files_struct(task);
1809 file = fcheck_files(files, fd);
1811 unsigned i_mode, f_mode = file->f_mode;
1814 put_files_struct(files);
1816 if (task_dumpable(task)) {
1818 cred = __task_cred(task);
1819 inode->i_uid = cred->euid;
1820 inode->i_gid = cred->egid;
1823 inode->i_uid = GLOBAL_ROOT_UID;
1824 inode->i_gid = GLOBAL_ROOT_GID;
1828 if (f_mode & FMODE_READ)
1829 i_mode |= S_IRUSR | S_IXUSR;
1830 if (f_mode & FMODE_WRITE)
1831 i_mode |= S_IWUSR | S_IXUSR;
1832 inode->i_mode = i_mode;
1834 security_task_to_inode(task, inode);
1835 put_task_struct(task);
1839 put_files_struct(files);
1841 put_task_struct(task);
1847 static const struct dentry_operations tid_fd_dentry_operations =
1849 .d_revalidate = tid_fd_revalidate,
1850 .d_delete = pid_delete_dentry,
1853 static struct dentry *proc_fd_instantiate(struct inode *dir,
1854 struct dentry *dentry, struct task_struct *task, const void *ptr)
1856 unsigned fd = *(const unsigned *)ptr;
1857 struct inode *inode;
1858 struct proc_inode *ei;
1859 struct dentry *error = ERR_PTR(-ENOENT);
1861 inode = proc_pid_make_inode(dir->i_sb, task);
1867 inode->i_op = &proc_pid_link_inode_operations;
1869 ei->op.proc_get_link = proc_fd_link;
1870 d_set_d_op(dentry, &tid_fd_dentry_operations);
1871 d_add(dentry, inode);
1872 /* Close the race of the process dying before we return the dentry */
1873 if (tid_fd_revalidate(dentry, NULL))
1880 static struct dentry *proc_lookupfd_common(struct inode *dir,
1881 struct dentry *dentry,
1882 instantiate_t instantiate)
1884 struct task_struct *task = get_proc_task(dir);
1885 unsigned fd = name_to_int(dentry);
1886 struct dentry *result = ERR_PTR(-ENOENT);
1893 result = instantiate(dir, dentry, task, &fd);
1895 put_task_struct(task);
1900 static int proc_readfd_common(struct file * filp, void * dirent,
1901 filldir_t filldir, instantiate_t instantiate)
1903 struct dentry *dentry = filp->f_path.dentry;
1904 struct inode *inode = dentry->d_inode;
1905 struct task_struct *p = get_proc_task(inode);
1906 unsigned int fd, ino;
1908 struct files_struct * files;
1918 if (filldir(dirent, ".", 1, 0, inode->i_ino, DT_DIR) < 0)
1922 ino = parent_ino(dentry);
1923 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
1927 files = get_files_struct(p);
1931 for (fd = filp->f_pos-2;
1932 fd < files_fdtable(files)->max_fds;
1933 fd++, filp->f_pos++) {
1934 char name[PROC_NUMBUF];
1937 if (!fcheck_files(files, fd))
1941 len = snprintf(name, sizeof(name), "%d", fd);
1942 if (proc_fill_cache(filp, dirent, filldir,
1943 name, len, instantiate,
1951 put_files_struct(files);
1959 static struct dentry *proc_lookupfd(struct inode *dir, struct dentry *dentry,
1960 struct nameidata *nd)
1962 return proc_lookupfd_common(dir, dentry, proc_fd_instantiate);
1965 static int proc_readfd(struct file *filp, void *dirent, filldir_t filldir)
1967 return proc_readfd_common(filp, dirent, filldir, proc_fd_instantiate);
1970 static ssize_t proc_fdinfo_read(struct file *file, char __user *buf,
1971 size_t len, loff_t *ppos)
1973 char tmp[PROC_FDINFO_MAX];
1974 int err = proc_fd_info(file->f_path.dentry->d_inode, NULL, tmp);
1976 err = simple_read_from_buffer(buf, len, ppos, tmp, strlen(tmp));
1980 static const struct file_operations proc_fdinfo_file_operations = {
1981 .open = nonseekable_open,
1982 .read = proc_fdinfo_read,
1983 .llseek = no_llseek,
1986 static const struct file_operations proc_fd_operations = {
1987 .read = generic_read_dir,
1988 .readdir = proc_readfd,
1989 .llseek = default_llseek,
1992 #ifdef CONFIG_CHECKPOINT_RESTORE
1995 * dname_to_vma_addr - maps a dentry name into two unsigned longs
1996 * which represent vma start and end addresses.
1998 static int dname_to_vma_addr(struct dentry *dentry,
1999 unsigned long *start, unsigned long *end)
2001 if (sscanf(dentry->d_name.name, "%lx-%lx", start, end) != 2)
2007 static int map_files_d_revalidate(struct dentry *dentry, struct nameidata *nd)
2009 unsigned long vm_start, vm_end;
2010 bool exact_vma_exists = false;
2011 struct mm_struct *mm = NULL;
2012 struct task_struct *task;
2013 const struct cred *cred;
2014 struct inode *inode;
2017 if (nd && nd->flags & LOOKUP_RCU)
2020 if (!capable(CAP_SYS_ADMIN)) {
2025 inode = dentry->d_inode;
2026 task = get_proc_task(inode);
2030 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2033 mm = get_task_mm(task);
2037 if (!dname_to_vma_addr(dentry, &vm_start, &vm_end)) {
2038 down_read(&mm->mmap_sem);
2039 exact_vma_exists = !!find_exact_vma(mm, vm_start, vm_end);
2040 up_read(&mm->mmap_sem);
2045 if (exact_vma_exists) {
2046 if (task_dumpable(task)) {
2048 cred = __task_cred(task);
2049 inode->i_uid = cred->euid;
2050 inode->i_gid = cred->egid;
2053 inode->i_uid = GLOBAL_ROOT_UID;
2054 inode->i_gid = GLOBAL_ROOT_GID;
2056 security_task_to_inode(task, inode);
2061 put_task_struct(task);
2070 static const struct dentry_operations tid_map_files_dentry_operations = {
2071 .d_revalidate = map_files_d_revalidate,
2072 .d_delete = pid_delete_dentry,
2075 static int proc_map_files_get_link(struct dentry *dentry, struct path *path)
2077 unsigned long vm_start, vm_end;
2078 struct vm_area_struct *vma;
2079 struct task_struct *task;
2080 struct mm_struct *mm;
2084 task = get_proc_task(dentry->d_inode);
2088 mm = get_task_mm(task);
2089 put_task_struct(task);
2093 rc = dname_to_vma_addr(dentry, &vm_start, &vm_end);
2097 down_read(&mm->mmap_sem);
2098 vma = find_exact_vma(mm, vm_start, vm_end);
2099 if (vma && vma->vm_file) {
2100 *path = vma->vm_file->f_path;
2104 up_read(&mm->mmap_sem);
2112 struct map_files_info {
2115 unsigned char name[4*sizeof(long)+2]; /* max: %lx-%lx\0 */
2118 static struct dentry *
2119 proc_map_files_instantiate(struct inode *dir, struct dentry *dentry,
2120 struct task_struct *task, const void *ptr)
2122 const struct file *file = ptr;
2123 struct proc_inode *ei;
2124 struct inode *inode;
2127 return ERR_PTR(-ENOENT);
2129 inode = proc_pid_make_inode(dir->i_sb, task);
2131 return ERR_PTR(-ENOENT);
2134 ei->op.proc_get_link = proc_map_files_get_link;
2136 inode->i_op = &proc_pid_link_inode_operations;
2138 inode->i_mode = S_IFLNK;
2140 if (file->f_mode & FMODE_READ)
2141 inode->i_mode |= S_IRUSR;
2142 if (file->f_mode & FMODE_WRITE)
2143 inode->i_mode |= S_IWUSR;
2145 d_set_d_op(dentry, &tid_map_files_dentry_operations);
2146 d_add(dentry, inode);
2151 static struct dentry *proc_map_files_lookup(struct inode *dir,
2152 struct dentry *dentry, struct nameidata *nd)
2154 unsigned long vm_start, vm_end;
2155 struct vm_area_struct *vma;
2156 struct task_struct *task;
2157 struct dentry *result;
2158 struct mm_struct *mm;
2160 result = ERR_PTR(-EACCES);
2161 if (!capable(CAP_SYS_ADMIN))
2164 result = ERR_PTR(-ENOENT);
2165 task = get_proc_task(dir);
2169 result = ERR_PTR(-EACCES);
2170 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2173 result = ERR_PTR(-ENOENT);
2174 if (dname_to_vma_addr(dentry, &vm_start, &vm_end))
2177 mm = get_task_mm(task);
2181 down_read(&mm->mmap_sem);
2182 vma = find_exact_vma(mm, vm_start, vm_end);
2186 result = proc_map_files_instantiate(dir, dentry, task, vma->vm_file);
2189 up_read(&mm->mmap_sem);
2192 put_task_struct(task);
2197 static const struct inode_operations proc_map_files_inode_operations = {
2198 .lookup = proc_map_files_lookup,
2199 .permission = proc_fd_permission,
2200 .setattr = proc_setattr,
2204 proc_map_files_readdir(struct file *filp, void *dirent, filldir_t filldir)
2206 struct dentry *dentry = filp->f_path.dentry;
2207 struct inode *inode = dentry->d_inode;
2208 struct vm_area_struct *vma;
2209 struct task_struct *task;
2210 struct mm_struct *mm;
2215 if (!capable(CAP_SYS_ADMIN))
2219 task = get_proc_task(inode);
2224 if (!ptrace_may_access(task, PTRACE_MODE_READ))
2228 switch (filp->f_pos) {
2231 if (filldir(dirent, ".", 1, 0, ino, DT_DIR) < 0)
2235 ino = parent_ino(dentry);
2236 if (filldir(dirent, "..", 2, 1, ino, DT_DIR) < 0)
2241 unsigned long nr_files, pos, i;
2242 struct flex_array *fa = NULL;
2243 struct map_files_info info;
2244 struct map_files_info *p;
2246 mm = get_task_mm(task);
2249 down_read(&mm->mmap_sem);
2254 * We need two passes here:
2256 * 1) Collect vmas of mapped files with mmap_sem taken
2257 * 2) Release mmap_sem and instantiate entries
2259 * otherwise we get lockdep complained, since filldir()
2260 * routine might require mmap_sem taken in might_fault().
2263 for (vma = mm->mmap, pos = 2; vma; vma = vma->vm_next) {
2264 if (vma->vm_file && ++pos > filp->f_pos)
2269 fa = flex_array_alloc(sizeof(info), nr_files,
2271 if (!fa || flex_array_prealloc(fa, 0, nr_files,
2275 flex_array_free(fa);
2276 up_read(&mm->mmap_sem);
2280 for (i = 0, vma = mm->mmap, pos = 2; vma;
2281 vma = vma->vm_next) {
2284 if (++pos <= filp->f_pos)
2287 get_file(vma->vm_file);
2288 info.file = vma->vm_file;
2289 info.len = snprintf(info.name,
2290 sizeof(info.name), "%lx-%lx",
2291 vma->vm_start, vma->vm_end);
2292 if (flex_array_put(fa, i++, &info, GFP_KERNEL))
2296 up_read(&mm->mmap_sem);
2298 for (i = 0; i < nr_files; i++) {
2299 p = flex_array_get(fa, i);
2300 ret = proc_fill_cache(filp, dirent, filldir,
2302 proc_map_files_instantiate,
2309 for (; i < nr_files; i++) {
2311 * In case of error don't forget
2312 * to put rest of file refs.
2314 p = flex_array_get(fa, i);
2318 flex_array_free(fa);
2324 put_task_struct(task);
2329 static const struct file_operations proc_map_files_operations = {
2330 .read = generic_read_dir,
2331 .readdir = proc_map_files_readdir,
2332 .llseek = default_llseek,
2335 #endif /* CONFIG_CHECKPOINT_RESTORE */
2338 * /proc/pid/fd needs a special permission handler so that a process can still
2339 * access /proc/self/fd after it has executed a setuid().
2341 static int proc_fd_permission(struct inode *inode, int mask)
2343 int rv = generic_permission(inode, mask);
2346 if (task_pid(current) == proc_pid(inode))
2352 * proc directories can do almost nothing..
2354 static const struct inode_operations proc_fd_inode_operations = {
2355 .lookup = proc_lookupfd,
2356 .permission = proc_fd_permission,
2357 .setattr = proc_setattr,
2360 static struct dentry *proc_fdinfo_instantiate(struct inode *dir,
2361 struct dentry *dentry, struct task_struct *task, const void *ptr)
2363 unsigned fd = *(unsigned *)ptr;
2364 struct inode *inode;
2365 struct proc_inode *ei;
2366 struct dentry *error = ERR_PTR(-ENOENT);
2368 inode = proc_pid_make_inode(dir->i_sb, task);
2373 inode->i_mode = S_IFREG | S_IRUSR;
2374 inode->i_fop = &proc_fdinfo_file_operations;
2375 d_set_d_op(dentry, &tid_fd_dentry_operations);
2376 d_add(dentry, inode);
2377 /* Close the race of the process dying before we return the dentry */
2378 if (tid_fd_revalidate(dentry, NULL))
2385 static struct dentry *proc_lookupfdinfo(struct inode *dir,
2386 struct dentry *dentry,
2387 struct nameidata *nd)
2389 return proc_lookupfd_common(dir, dentry, proc_fdinfo_instantiate);
2392 static int proc_readfdinfo(struct file *filp, void *dirent, filldir_t filldir)
2394 return proc_readfd_common(filp, dirent, filldir,
2395 proc_fdinfo_instantiate);
2398 static const struct file_operations proc_fdinfo_operations = {
2399 .read = generic_read_dir,
2400 .readdir = proc_readfdinfo,
2401 .llseek = default_llseek,
2405 * proc directories can do almost nothing..
2407 static const struct inode_operations proc_fdinfo_inode_operations = {
2408 .lookup = proc_lookupfdinfo,
2409 .setattr = proc_setattr,
2413 static struct dentry *proc_pident_instantiate(struct inode *dir,
2414 struct dentry *dentry, struct task_struct *task, const void *ptr)
2416 const struct pid_entry *p = ptr;
2417 struct inode *inode;
2418 struct proc_inode *ei;
2419 struct dentry *error = ERR_PTR(-ENOENT);
2421 inode = proc_pid_make_inode(dir->i_sb, task);
2426 inode->i_mode = p->mode;
2427 if (S_ISDIR(inode->i_mode))
2428 set_nlink(inode, 2); /* Use getattr to fix if necessary */
2430 inode->i_op = p->iop;
2432 inode->i_fop = p->fop;
2434 d_set_d_op(dentry, &pid_dentry_operations);
2435 d_add(dentry, inode);
2436 /* Close the race of the process dying before we return the dentry */
2437 if (pid_revalidate(dentry, NULL))
2443 static struct dentry *proc_pident_lookup(struct inode *dir,
2444 struct dentry *dentry,
2445 const struct pid_entry *ents,
2448 struct dentry *error;
2449 struct task_struct *task = get_proc_task(dir);
2450 const struct pid_entry *p, *last;
2452 error = ERR_PTR(-ENOENT);
2458 * Yes, it does not scale. And it should not. Don't add
2459 * new entries into /proc/<tgid>/ without very good reasons.
2461 last = &ents[nents - 1];
2462 for (p = ents; p <= last; p++) {
2463 if (p->len != dentry->d_name.len)
2465 if (!memcmp(dentry->d_name.name, p->name, p->len))
2471 error = proc_pident_instantiate(dir, dentry, task, p);
2473 put_task_struct(task);
2478 static int proc_pident_fill_cache(struct file *filp, void *dirent,
2479 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2481 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2482 proc_pident_instantiate, task, p);
2485 static int proc_pident_readdir(struct file *filp,
2486 void *dirent, filldir_t filldir,
2487 const struct pid_entry *ents, unsigned int nents)
2490 struct dentry *dentry = filp->f_path.dentry;
2491 struct inode *inode = dentry->d_inode;
2492 struct task_struct *task = get_proc_task(inode);
2493 const struct pid_entry *p, *last;
2506 if (filldir(dirent, ".", 1, i, ino, DT_DIR) < 0)
2512 ino = parent_ino(dentry);
2513 if (filldir(dirent, "..", 2, i, ino, DT_DIR) < 0)
2525 last = &ents[nents - 1];
2527 if (proc_pident_fill_cache(filp, dirent, filldir, task, p) < 0)
2536 put_task_struct(task);
2541 #ifdef CONFIG_SECURITY
2542 static ssize_t proc_pid_attr_read(struct file * file, char __user * buf,
2543 size_t count, loff_t *ppos)
2545 struct inode * inode = file->f_path.dentry->d_inode;
2548 struct task_struct *task = get_proc_task(inode);
2553 length = security_getprocattr(task,
2554 (char*)file->f_path.dentry->d_name.name,
2556 put_task_struct(task);
2558 length = simple_read_from_buffer(buf, count, ppos, p, length);
2563 static ssize_t proc_pid_attr_write(struct file * file, const char __user * buf,
2564 size_t count, loff_t *ppos)
2566 struct inode * inode = file->f_path.dentry->d_inode;
2569 struct task_struct *task = get_proc_task(inode);
2574 if (count > PAGE_SIZE)
2577 /* No partial writes. */
2583 page = (char*)__get_free_page(GFP_TEMPORARY);
2588 if (copy_from_user(page, buf, count))
2591 /* Guard against adverse ptrace interaction */
2592 length = mutex_lock_interruptible(&task->signal->cred_guard_mutex);
2596 length = security_setprocattr(task,
2597 (char*)file->f_path.dentry->d_name.name,
2598 (void*)page, count);
2599 mutex_unlock(&task->signal->cred_guard_mutex);
2601 free_page((unsigned long) page);
2603 put_task_struct(task);
2608 static const struct file_operations proc_pid_attr_operations = {
2609 .read = proc_pid_attr_read,
2610 .write = proc_pid_attr_write,
2611 .llseek = generic_file_llseek,
2614 static const struct pid_entry attr_dir_stuff[] = {
2615 REG("current", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2616 REG("prev", S_IRUGO, proc_pid_attr_operations),
2617 REG("exec", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2618 REG("fscreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2619 REG("keycreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2620 REG("sockcreate", S_IRUGO|S_IWUGO, proc_pid_attr_operations),
2623 static int proc_attr_dir_readdir(struct file * filp,
2624 void * dirent, filldir_t filldir)
2626 return proc_pident_readdir(filp,dirent,filldir,
2627 attr_dir_stuff,ARRAY_SIZE(attr_dir_stuff));
2630 static const struct file_operations proc_attr_dir_operations = {
2631 .read = generic_read_dir,
2632 .readdir = proc_attr_dir_readdir,
2633 .llseek = default_llseek,
2636 static struct dentry *proc_attr_dir_lookup(struct inode *dir,
2637 struct dentry *dentry, struct nameidata *nd)
2639 return proc_pident_lookup(dir, dentry,
2640 attr_dir_stuff, ARRAY_SIZE(attr_dir_stuff));
2643 static const struct inode_operations proc_attr_dir_inode_operations = {
2644 .lookup = proc_attr_dir_lookup,
2645 .getattr = pid_getattr,
2646 .setattr = proc_setattr,
2651 #ifdef CONFIG_ELF_CORE
2652 static ssize_t proc_coredump_filter_read(struct file *file, char __user *buf,
2653 size_t count, loff_t *ppos)
2655 struct task_struct *task = get_proc_task(file->f_dentry->d_inode);
2656 struct mm_struct *mm;
2657 char buffer[PROC_NUMBUF];
2665 mm = get_task_mm(task);
2667 len = snprintf(buffer, sizeof(buffer), "%08lx\n",
2668 ((mm->flags & MMF_DUMP_FILTER_MASK) >>
2669 MMF_DUMP_FILTER_SHIFT));
2671 ret = simple_read_from_buffer(buf, count, ppos, buffer, len);
2674 put_task_struct(task);
2679 static ssize_t proc_coredump_filter_write(struct file *file,
2680 const char __user *buf,
2684 struct task_struct *task;
2685 struct mm_struct *mm;
2686 char buffer[PROC_NUMBUF], *end;
2693 memset(buffer, 0, sizeof(buffer));
2694 if (count > sizeof(buffer) - 1)
2695 count = sizeof(buffer) - 1;
2696 if (copy_from_user(buffer, buf, count))
2700 val = (unsigned int)simple_strtoul(buffer, &end, 0);
2703 if (end - buffer == 0)
2707 task = get_proc_task(file->f_dentry->d_inode);
2712 mm = get_task_mm(task);
2716 for (i = 0, mask = 1; i < MMF_DUMP_FILTER_BITS; i++, mask <<= 1) {
2718 set_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2720 clear_bit(i + MMF_DUMP_FILTER_SHIFT, &mm->flags);
2725 put_task_struct(task);
2730 static const struct file_operations proc_coredump_filter_operations = {
2731 .read = proc_coredump_filter_read,
2732 .write = proc_coredump_filter_write,
2733 .llseek = generic_file_llseek,
2740 static int proc_self_readlink(struct dentry *dentry, char __user *buffer,
2743 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2744 pid_t tgid = task_tgid_nr_ns(current, ns);
2745 char tmp[PROC_NUMBUF];
2748 sprintf(tmp, "%d", tgid);
2749 return vfs_readlink(dentry,buffer,buflen,tmp);
2752 static void *proc_self_follow_link(struct dentry *dentry, struct nameidata *nd)
2754 struct pid_namespace *ns = dentry->d_sb->s_fs_info;
2755 pid_t tgid = task_tgid_nr_ns(current, ns);
2756 char *name = ERR_PTR(-ENOENT);
2760 name = ERR_PTR(-ENOMEM);
2762 sprintf(name, "%d", tgid);
2764 nd_set_link(nd, name);
2768 static void proc_self_put_link(struct dentry *dentry, struct nameidata *nd,
2771 char *s = nd_get_link(nd);
2776 static const struct inode_operations proc_self_inode_operations = {
2777 .readlink = proc_self_readlink,
2778 .follow_link = proc_self_follow_link,
2779 .put_link = proc_self_put_link,
2785 * These are the directory entries in the root directory of /proc
2786 * that properly belong to the /proc filesystem, as they describe
2787 * describe something that is process related.
2789 static const struct pid_entry proc_base_stuff[] = {
2790 NOD("self", S_IFLNK|S_IRWXUGO,
2791 &proc_self_inode_operations, NULL, {}),
2794 static struct dentry *proc_base_instantiate(struct inode *dir,
2795 struct dentry *dentry, struct task_struct *task, const void *ptr)
2797 const struct pid_entry *p = ptr;
2798 struct inode *inode;
2799 struct proc_inode *ei;
2800 struct dentry *error;
2802 /* Allocate the inode */
2803 error = ERR_PTR(-ENOMEM);
2804 inode = new_inode(dir->i_sb);
2808 /* Initialize the inode */
2810 inode->i_ino = get_next_ino();
2811 inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
2814 * grab the reference to the task.
2816 ei->pid = get_task_pid(task, PIDTYPE_PID);
2820 inode->i_mode = p->mode;
2821 if (S_ISDIR(inode->i_mode))
2822 set_nlink(inode, 2);
2823 if (S_ISLNK(inode->i_mode))
2826 inode->i_op = p->iop;
2828 inode->i_fop = p->fop;
2830 d_add(dentry, inode);
2839 static struct dentry *proc_base_lookup(struct inode *dir, struct dentry *dentry)
2841 struct dentry *error;
2842 struct task_struct *task = get_proc_task(dir);
2843 const struct pid_entry *p, *last;
2845 error = ERR_PTR(-ENOENT);
2850 /* Lookup the directory entry */
2851 last = &proc_base_stuff[ARRAY_SIZE(proc_base_stuff) - 1];
2852 for (p = proc_base_stuff; p <= last; p++) {
2853 if (p->len != dentry->d_name.len)
2855 if (!memcmp(dentry->d_name.name, p->name, p->len))
2861 error = proc_base_instantiate(dir, dentry, task, p);
2864 put_task_struct(task);
2869 static int proc_base_fill_cache(struct file *filp, void *dirent,
2870 filldir_t filldir, struct task_struct *task, const struct pid_entry *p)
2872 return proc_fill_cache(filp, dirent, filldir, p->name, p->len,
2873 proc_base_instantiate, task, p);
2876 #ifdef CONFIG_TASK_IO_ACCOUNTING
2877 static int do_io_accounting(struct task_struct *task, char *buffer, int whole)
2879 struct task_io_accounting acct = task->ioac;
2880 unsigned long flags;
2883 result = mutex_lock_killable(&task->signal->cred_guard_mutex);
2887 if (!ptrace_may_access(task, PTRACE_MODE_READ)) {
2892 if (whole && lock_task_sighand(task, &flags)) {
2893 struct task_struct *t = task;
2895 task_io_accounting_add(&acct, &task->signal->ioac);
2896 while_each_thread(task, t)
2897 task_io_accounting_add(&acct, &t->ioac);
2899 unlock_task_sighand(task, &flags);
2901 result = sprintf(buffer,
2906 "read_bytes: %llu\n"
2907 "write_bytes: %llu\n"
2908 "cancelled_write_bytes: %llu\n",
2909 (unsigned long long)acct.rchar,
2910 (unsigned long long)acct.wchar,
2911 (unsigned long long)acct.syscr,
2912 (unsigned long long)acct.syscw,
2913 (unsigned long long)acct.read_bytes,
2914 (unsigned long long)acct.write_bytes,
2915 (unsigned long long)acct.cancelled_write_bytes);
2917 mutex_unlock(&task->signal->cred_guard_mutex);
2921 static int proc_tid_io_accounting(struct task_struct *task, char *buffer)
2923 return do_io_accounting(task, buffer, 0);
2926 static int proc_tgid_io_accounting(struct task_struct *task, char *buffer)
2928 return do_io_accounting(task, buffer, 1);
2930 #endif /* CONFIG_TASK_IO_ACCOUNTING */
2932 #ifdef CONFIG_USER_NS
2933 static int proc_id_map_open(struct inode *inode, struct file *file,
2934 struct seq_operations *seq_ops)
2936 struct user_namespace *ns = NULL;
2937 struct task_struct *task;
2938 struct seq_file *seq;
2941 task = get_proc_task(inode);
2944 ns = get_user_ns(task_cred_xxx(task, user_ns));
2946 put_task_struct(task);
2951 ret = seq_open(file, seq_ops);
2955 seq = file->private_data;
2965 static int proc_id_map_release(struct inode *inode, struct file *file)
2967 struct seq_file *seq = file->private_data;
2968 struct user_namespace *ns = seq->private;
2970 return seq_release(inode, file);
2973 static int proc_uid_map_open(struct inode *inode, struct file *file)
2975 return proc_id_map_open(inode, file, &proc_uid_seq_operations);
2978 static int proc_gid_map_open(struct inode *inode, struct file *file)
2980 return proc_id_map_open(inode, file, &proc_gid_seq_operations);
2983 static const struct file_operations proc_uid_map_operations = {
2984 .open = proc_uid_map_open,
2985 .write = proc_uid_map_write,
2987 .llseek = seq_lseek,
2988 .release = proc_id_map_release,
2991 static const struct file_operations proc_gid_map_operations = {
2992 .open = proc_gid_map_open,
2993 .write = proc_gid_map_write,
2995 .llseek = seq_lseek,
2996 .release = proc_id_map_release,
2998 #endif /* CONFIG_USER_NS */
3000 static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
3001 struct pid *pid, struct task_struct *task)
3003 int err = lock_trace(task);
3005 seq_printf(m, "%08x\n", task->personality);
3014 static const struct file_operations proc_task_operations;
3015 static const struct inode_operations proc_task_inode_operations;
3017 static const struct pid_entry tgid_base_stuff[] = {
3018 DIR("task", S_IRUGO|S_IXUGO, proc_task_inode_operations, proc_task_operations),
3019 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3020 #ifdef CONFIG_CHECKPOINT_RESTORE
3021 DIR("map_files", S_IRUSR|S_IXUSR, proc_map_files_inode_operations, proc_map_files_operations),
3023 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3024 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3026 DIR("net", S_IRUGO|S_IXUGO, proc_net_inode_operations, proc_net_operations),
3028 REG("environ", S_IRUSR, proc_environ_operations),
3029 INF("auxv", S_IRUSR, proc_pid_auxv),
3030 ONE("status", S_IRUGO, proc_pid_status),
3031 ONE("personality", S_IRUGO, proc_pid_personality),
3032 INF("limits", S_IRUGO, proc_pid_limits),
3033 #ifdef CONFIG_SCHED_DEBUG
3034 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3036 #ifdef CONFIG_SCHED_AUTOGROUP
3037 REG("autogroup", S_IRUGO|S_IWUSR, proc_pid_sched_autogroup_operations),
3039 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3040 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3041 INF("syscall", S_IRUGO, proc_pid_syscall),
3043 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3044 ONE("stat", S_IRUGO, proc_tgid_stat),
3045 ONE("statm", S_IRUGO, proc_pid_statm),
3046 REG("maps", S_IRUGO, proc_pid_maps_operations),
3048 REG("numa_maps", S_IRUGO, proc_pid_numa_maps_operations),
3050 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3051 LNK("cwd", proc_cwd_link),
3052 LNK("root", proc_root_link),
3053 LNK("exe", proc_exe_link),
3054 REG("mounts", S_IRUGO, proc_mounts_operations),
3055 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3056 REG("mountstats", S_IRUSR, proc_mountstats_operations),
3057 #ifdef CONFIG_PROC_PAGE_MONITOR
3058 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3059 REG("smaps", S_IRUGO, proc_pid_smaps_operations),
3060 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3062 #ifdef CONFIG_SECURITY
3063 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3065 #ifdef CONFIG_KALLSYMS
3066 INF("wchan", S_IRUGO, proc_pid_wchan),
3068 #ifdef CONFIG_STACKTRACE
3069 ONE("stack", S_IRUGO, proc_pid_stack),
3071 #ifdef CONFIG_SCHEDSTATS
3072 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3074 #ifdef CONFIG_LATENCYTOP
3075 REG("latency", S_IRUGO, proc_lstats_operations),
3077 #ifdef CONFIG_PROC_PID_CPUSET
3078 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3080 #ifdef CONFIG_CGROUPS
3081 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3083 INF("oom_score", S_IRUGO, proc_oom_score),
3084 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3085 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3086 #ifdef CONFIG_AUDITSYSCALL
3087 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3088 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3090 #ifdef CONFIG_FAULT_INJECTION
3091 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3093 #ifdef CONFIG_ELF_CORE
3094 REG("coredump_filter", S_IRUGO|S_IWUSR, proc_coredump_filter_operations),
3096 #ifdef CONFIG_TASK_IO_ACCOUNTING
3097 INF("io", S_IRUSR, proc_tgid_io_accounting),
3099 #ifdef CONFIG_HARDWALL
3100 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3102 #ifdef CONFIG_USER_NS
3103 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3104 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3108 static int proc_tgid_base_readdir(struct file * filp,
3109 void * dirent, filldir_t filldir)
3111 return proc_pident_readdir(filp,dirent,filldir,
3112 tgid_base_stuff,ARRAY_SIZE(tgid_base_stuff));
3115 static const struct file_operations proc_tgid_base_operations = {
3116 .read = generic_read_dir,
3117 .readdir = proc_tgid_base_readdir,
3118 .llseek = default_llseek,
3121 static struct dentry *proc_tgid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3122 return proc_pident_lookup(dir, dentry,
3123 tgid_base_stuff, ARRAY_SIZE(tgid_base_stuff));
3126 static const struct inode_operations proc_tgid_base_inode_operations = {
3127 .lookup = proc_tgid_base_lookup,
3128 .getattr = pid_getattr,
3129 .setattr = proc_setattr,
3130 .permission = proc_pid_permission,
3133 static void proc_flush_task_mnt(struct vfsmount *mnt, pid_t pid, pid_t tgid)
3135 struct dentry *dentry, *leader, *dir;
3136 char buf[PROC_NUMBUF];
3140 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3141 dentry = d_hash_and_lookup(mnt->mnt_root, &name);
3143 shrink_dcache_parent(dentry);
3149 name.len = snprintf(buf, sizeof(buf), "%d", tgid);
3150 leader = d_hash_and_lookup(mnt->mnt_root, &name);
3155 name.len = strlen(name.name);
3156 dir = d_hash_and_lookup(leader, &name);
3158 goto out_put_leader;
3161 name.len = snprintf(buf, sizeof(buf), "%d", pid);
3162 dentry = d_hash_and_lookup(dir, &name);
3164 shrink_dcache_parent(dentry);
3177 * proc_flush_task - Remove dcache entries for @task from the /proc dcache.
3178 * @task: task that should be flushed.
3180 * When flushing dentries from proc, one needs to flush them from global
3181 * proc (proc_mnt) and from all the namespaces' procs this task was seen
3182 * in. This call is supposed to do all of this job.
3184 * Looks in the dcache for
3186 * /proc/@tgid/task/@pid
3187 * if either directory is present flushes it and all of it'ts children
3190 * It is safe and reasonable to cache /proc entries for a task until
3191 * that task exits. After that they just clog up the dcache with
3192 * useless entries, possibly causing useful dcache entries to be
3193 * flushed instead. This routine is proved to flush those useless
3194 * dcache entries at process exit time.
3196 * NOTE: This routine is just an optimization so it does not guarantee
3197 * that no dcache entries will exist at process exit time it
3198 * just makes it very unlikely that any will persist.
3201 void proc_flush_task(struct task_struct *task)
3204 struct pid *pid, *tgid;
3207 pid = task_pid(task);
3208 tgid = task_tgid(task);
3210 for (i = 0; i <= pid->level; i++) {
3211 upid = &pid->numbers[i];
3212 proc_flush_task_mnt(upid->ns->proc_mnt, upid->nr,
3213 tgid->numbers[i].nr);
3216 upid = &pid->numbers[pid->level];
3218 pid_ns_release_proc(upid->ns);
3221 static struct dentry *proc_pid_instantiate(struct inode *dir,
3222 struct dentry * dentry,
3223 struct task_struct *task, const void *ptr)
3225 struct dentry *error = ERR_PTR(-ENOENT);
3226 struct inode *inode;
3228 inode = proc_pid_make_inode(dir->i_sb, task);
3232 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3233 inode->i_op = &proc_tgid_base_inode_operations;
3234 inode->i_fop = &proc_tgid_base_operations;
3235 inode->i_flags|=S_IMMUTABLE;
3237 set_nlink(inode, 2 + pid_entry_count_dirs(tgid_base_stuff,
3238 ARRAY_SIZE(tgid_base_stuff)));
3240 d_set_d_op(dentry, &pid_dentry_operations);
3242 d_add(dentry, inode);
3243 /* Close the race of the process dying before we return the dentry */
3244 if (pid_revalidate(dentry, NULL))
3250 struct dentry *proc_pid_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3252 struct dentry *result;
3253 struct task_struct *task;
3255 struct pid_namespace *ns;
3257 result = proc_base_lookup(dir, dentry);
3258 if (!IS_ERR(result) || PTR_ERR(result) != -ENOENT)
3261 tgid = name_to_int(dentry);
3265 ns = dentry->d_sb->s_fs_info;
3267 task = find_task_by_pid_ns(tgid, ns);
3269 get_task_struct(task);
3274 result = proc_pid_instantiate(dir, dentry, task, NULL);
3275 put_task_struct(task);
3281 * Find the first task with tgid >= tgid
3286 struct task_struct *task;
3288 static struct tgid_iter next_tgid(struct pid_namespace *ns, struct tgid_iter iter)
3293 put_task_struct(iter.task);
3297 pid = find_ge_pid(iter.tgid, ns);
3299 iter.tgid = pid_nr_ns(pid, ns);
3300 iter.task = pid_task(pid, PIDTYPE_PID);
3301 /* What we to know is if the pid we have find is the
3302 * pid of a thread_group_leader. Testing for task
3303 * being a thread_group_leader is the obvious thing
3304 * todo but there is a window when it fails, due to
3305 * the pid transfer logic in de_thread.
3307 * So we perform the straight forward test of seeing
3308 * if the pid we have found is the pid of a thread
3309 * group leader, and don't worry if the task we have
3310 * found doesn't happen to be a thread group leader.
3311 * As we don't care in the case of readdir.
3313 if (!iter.task || !has_group_leader_pid(iter.task)) {
3317 get_task_struct(iter.task);
3323 #define TGID_OFFSET (FIRST_PROCESS_ENTRY + ARRAY_SIZE(proc_base_stuff))
3325 static int proc_pid_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3326 struct tgid_iter iter)
3328 char name[PROC_NUMBUF];
3329 int len = snprintf(name, sizeof(name), "%d", iter.tgid);
3330 return proc_fill_cache(filp, dirent, filldir, name, len,
3331 proc_pid_instantiate, iter.task, NULL);
3334 static int fake_filldir(void *buf, const char *name, int namelen,
3335 loff_t offset, u64 ino, unsigned d_type)
3340 /* for the /proc/ directory itself, after non-process stuff has been done */
3341 int proc_pid_readdir(struct file * filp, void * dirent, filldir_t filldir)
3344 struct task_struct *reaper;
3345 struct tgid_iter iter;
3346 struct pid_namespace *ns;
3347 filldir_t __filldir;
3349 if (filp->f_pos >= PID_MAX_LIMIT + TGID_OFFSET)
3351 nr = filp->f_pos - FIRST_PROCESS_ENTRY;
3353 reaper = get_proc_task(filp->f_path.dentry->d_inode);
3357 for (; nr < ARRAY_SIZE(proc_base_stuff); filp->f_pos++, nr++) {
3358 const struct pid_entry *p = &proc_base_stuff[nr];
3359 if (proc_base_fill_cache(filp, dirent, filldir, reaper, p) < 0)
3363 ns = filp->f_dentry->d_sb->s_fs_info;
3365 iter.tgid = filp->f_pos - TGID_OFFSET;
3366 for (iter = next_tgid(ns, iter);
3368 iter.tgid += 1, iter = next_tgid(ns, iter)) {
3369 if (has_pid_permissions(ns, iter.task, 2))
3370 __filldir = filldir;
3372 __filldir = fake_filldir;
3374 filp->f_pos = iter.tgid + TGID_OFFSET;
3375 if (proc_pid_fill_cache(filp, dirent, __filldir, iter) < 0) {
3376 put_task_struct(iter.task);
3380 filp->f_pos = PID_MAX_LIMIT + TGID_OFFSET;
3382 put_task_struct(reaper);
3390 static const struct pid_entry tid_base_stuff[] = {
3391 DIR("fd", S_IRUSR|S_IXUSR, proc_fd_inode_operations, proc_fd_operations),
3392 DIR("fdinfo", S_IRUSR|S_IXUSR, proc_fdinfo_inode_operations, proc_fdinfo_operations),
3393 DIR("ns", S_IRUSR|S_IXUGO, proc_ns_dir_inode_operations, proc_ns_dir_operations),
3394 REG("environ", S_IRUSR, proc_environ_operations),
3395 INF("auxv", S_IRUSR, proc_pid_auxv),
3396 ONE("status", S_IRUGO, proc_pid_status),
3397 ONE("personality", S_IRUGO, proc_pid_personality),
3398 INF("limits", S_IRUGO, proc_pid_limits),
3399 #ifdef CONFIG_SCHED_DEBUG
3400 REG("sched", S_IRUGO|S_IWUSR, proc_pid_sched_operations),
3402 REG("comm", S_IRUGO|S_IWUSR, proc_pid_set_comm_operations),
3403 #ifdef CONFIG_HAVE_ARCH_TRACEHOOK
3404 INF("syscall", S_IRUGO, proc_pid_syscall),
3406 INF("cmdline", S_IRUGO, proc_pid_cmdline),
3407 ONE("stat", S_IRUGO, proc_tid_stat),
3408 ONE("statm", S_IRUGO, proc_pid_statm),
3409 REG("maps", S_IRUGO, proc_tid_maps_operations),
3411 REG("numa_maps", S_IRUGO, proc_tid_numa_maps_operations),
3413 REG("mem", S_IRUSR|S_IWUSR, proc_mem_operations),
3414 LNK("cwd", proc_cwd_link),
3415 LNK("root", proc_root_link),
3416 LNK("exe", proc_exe_link),
3417 REG("mounts", S_IRUGO, proc_mounts_operations),
3418 REG("mountinfo", S_IRUGO, proc_mountinfo_operations),
3419 #ifdef CONFIG_PROC_PAGE_MONITOR
3420 REG("clear_refs", S_IWUSR, proc_clear_refs_operations),
3421 REG("smaps", S_IRUGO, proc_tid_smaps_operations),
3422 REG("pagemap", S_IRUGO, proc_pagemap_operations),
3424 #ifdef CONFIG_SECURITY
3425 DIR("attr", S_IRUGO|S_IXUGO, proc_attr_dir_inode_operations, proc_attr_dir_operations),
3427 #ifdef CONFIG_KALLSYMS
3428 INF("wchan", S_IRUGO, proc_pid_wchan),
3430 #ifdef CONFIG_STACKTRACE
3431 ONE("stack", S_IRUGO, proc_pid_stack),
3433 #ifdef CONFIG_SCHEDSTATS
3434 INF("schedstat", S_IRUGO, proc_pid_schedstat),
3436 #ifdef CONFIG_LATENCYTOP
3437 REG("latency", S_IRUGO, proc_lstats_operations),
3439 #ifdef CONFIG_PROC_PID_CPUSET
3440 REG("cpuset", S_IRUGO, proc_cpuset_operations),
3442 #ifdef CONFIG_CGROUPS
3443 REG("cgroup", S_IRUGO, proc_cgroup_operations),
3445 INF("oom_score", S_IRUGO, proc_oom_score),
3446 REG("oom_adj", S_IRUGO|S_IWUSR, proc_oom_adjust_operations),
3447 REG("oom_score_adj", S_IRUGO|S_IWUSR, proc_oom_score_adj_operations),
3448 #ifdef CONFIG_AUDITSYSCALL
3449 REG("loginuid", S_IWUSR|S_IRUGO, proc_loginuid_operations),
3450 REG("sessionid", S_IRUGO, proc_sessionid_operations),
3452 #ifdef CONFIG_FAULT_INJECTION
3453 REG("make-it-fail", S_IRUGO|S_IWUSR, proc_fault_inject_operations),
3455 #ifdef CONFIG_TASK_IO_ACCOUNTING
3456 INF("io", S_IRUSR, proc_tid_io_accounting),
3458 #ifdef CONFIG_HARDWALL
3459 INF("hardwall", S_IRUGO, proc_pid_hardwall),
3461 #ifdef CONFIG_USER_NS
3462 REG("uid_map", S_IRUGO|S_IWUSR, proc_uid_map_operations),
3463 REG("gid_map", S_IRUGO|S_IWUSR, proc_gid_map_operations),
3467 static int proc_tid_base_readdir(struct file * filp,
3468 void * dirent, filldir_t filldir)
3470 return proc_pident_readdir(filp,dirent,filldir,
3471 tid_base_stuff,ARRAY_SIZE(tid_base_stuff));
3474 static struct dentry *proc_tid_base_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd){
3475 return proc_pident_lookup(dir, dentry,
3476 tid_base_stuff, ARRAY_SIZE(tid_base_stuff));
3479 static const struct file_operations proc_tid_base_operations = {
3480 .read = generic_read_dir,
3481 .readdir = proc_tid_base_readdir,
3482 .llseek = default_llseek,
3485 static const struct inode_operations proc_tid_base_inode_operations = {
3486 .lookup = proc_tid_base_lookup,
3487 .getattr = pid_getattr,
3488 .setattr = proc_setattr,
3491 static struct dentry *proc_task_instantiate(struct inode *dir,
3492 struct dentry *dentry, struct task_struct *task, const void *ptr)
3494 struct dentry *error = ERR_PTR(-ENOENT);
3495 struct inode *inode;
3496 inode = proc_pid_make_inode(dir->i_sb, task);
3500 inode->i_mode = S_IFDIR|S_IRUGO|S_IXUGO;
3501 inode->i_op = &proc_tid_base_inode_operations;
3502 inode->i_fop = &proc_tid_base_operations;
3503 inode->i_flags|=S_IMMUTABLE;
3505 set_nlink(inode, 2 + pid_entry_count_dirs(tid_base_stuff,
3506 ARRAY_SIZE(tid_base_stuff)));
3508 d_set_d_op(dentry, &pid_dentry_operations);
3510 d_add(dentry, inode);
3511 /* Close the race of the process dying before we return the dentry */
3512 if (pid_revalidate(dentry, NULL))
3518 static struct dentry *proc_task_lookup(struct inode *dir, struct dentry * dentry, struct nameidata *nd)
3520 struct dentry *result = ERR_PTR(-ENOENT);
3521 struct task_struct *task;
3522 struct task_struct *leader = get_proc_task(dir);
3524 struct pid_namespace *ns;
3529 tid = name_to_int(dentry);
3533 ns = dentry->d_sb->s_fs_info;
3535 task = find_task_by_pid_ns(tid, ns);
3537 get_task_struct(task);
3541 if (!same_thread_group(leader, task))
3544 result = proc_task_instantiate(dir, dentry, task, NULL);
3546 put_task_struct(task);
3548 put_task_struct(leader);
3554 * Find the first tid of a thread group to return to user space.
3556 * Usually this is just the thread group leader, but if the users
3557 * buffer was too small or there was a seek into the middle of the
3558 * directory we have more work todo.
3560 * In the case of a short read we start with find_task_by_pid.
3562 * In the case of a seek we start with the leader and walk nr
3565 static struct task_struct *first_tid(struct task_struct *leader,
3566 int tid, int nr, struct pid_namespace *ns)
3568 struct task_struct *pos;
3571 /* Attempt to start with the pid of a thread */
3572 if (tid && (nr > 0)) {
3573 pos = find_task_by_pid_ns(tid, ns);
3574 if (pos && (pos->group_leader == leader))
3578 /* If nr exceeds the number of threads there is nothing todo */
3580 if (nr && nr >= get_nr_threads(leader))
3583 /* If we haven't found our starting place yet start
3584 * with the leader and walk nr threads forward.
3586 for (pos = leader; nr > 0; --nr) {
3587 pos = next_thread(pos);
3588 if (pos == leader) {
3594 get_task_struct(pos);
3601 * Find the next thread in the thread list.
3602 * Return NULL if there is an error or no next thread.
3604 * The reference to the input task_struct is released.
3606 static struct task_struct *next_tid(struct task_struct *start)
3608 struct task_struct *pos = NULL;
3610 if (pid_alive(start)) {
3611 pos = next_thread(start);
3612 if (thread_group_leader(pos))
3615 get_task_struct(pos);
3618 put_task_struct(start);
3622 static int proc_task_fill_cache(struct file *filp, void *dirent, filldir_t filldir,
3623 struct task_struct *task, int tid)
3625 char name[PROC_NUMBUF];
3626 int len = snprintf(name, sizeof(name), "%d", tid);
3627 return proc_fill_cache(filp, dirent, filldir, name, len,
3628 proc_task_instantiate, task, NULL);
3631 /* for the /proc/TGID/task/ directories */
3632 static int proc_task_readdir(struct file * filp, void * dirent, filldir_t filldir)
3634 struct dentry *dentry = filp->f_path.dentry;
3635 struct inode *inode = dentry->d_inode;
3636 struct task_struct *leader = NULL;
3637 struct task_struct *task;
3638 int retval = -ENOENT;
3641 struct pid_namespace *ns;
3643 task = get_proc_task(inode);
3647 if (pid_alive(task)) {
3648 leader = task->group_leader;
3649 get_task_struct(leader);
3652 put_task_struct(task);
3657 switch ((unsigned long)filp->f_pos) {
3660 if (filldir(dirent, ".", 1, filp->f_pos, ino, DT_DIR) < 0)
3665 ino = parent_ino(dentry);
3666 if (filldir(dirent, "..", 2, filp->f_pos, ino, DT_DIR) < 0)
3672 /* f_version caches the tgid value that the last readdir call couldn't
3673 * return. lseek aka telldir automagically resets f_version to 0.
3675 ns = filp->f_dentry->d_sb->s_fs_info;
3676 tid = (int)filp->f_version;
3677 filp->f_version = 0;
3678 for (task = first_tid(leader, tid, filp->f_pos - 2, ns);
3680 task = next_tid(task), filp->f_pos++) {
3681 tid = task_pid_nr_ns(task, ns);
3682 if (proc_task_fill_cache(filp, dirent, filldir, task, tid) < 0) {
3683 /* returning this tgid failed, save it as the first
3684 * pid for the next readir call */
3685 filp->f_version = (u64)tid;
3686 put_task_struct(task);
3691 put_task_struct(leader);
3696 static int proc_task_getattr(struct vfsmount *mnt, struct dentry *dentry, struct kstat *stat)
3698 struct inode *inode = dentry->d_inode;
3699 struct task_struct *p = get_proc_task(inode);
3700 generic_fillattr(inode, stat);
3703 stat->nlink += get_nr_threads(p);
3710 static const struct inode_operations proc_task_inode_operations = {
3711 .lookup = proc_task_lookup,
3712 .getattr = proc_task_getattr,
3713 .setattr = proc_setattr,
3714 .permission = proc_pid_permission,
3717 static const struct file_operations proc_task_operations = {
3718 .read = generic_read_dir,
3719 .readdir = proc_task_readdir,
3720 .llseek = default_llseek,